Skip to content

BusinessPractitioner2 min read

How much does faster recovery pay off?

What are faster detection and restoration worth? Incident cost before and after, return on investment, payback period and sensitivity to assumptions.

For: team managers · finance, risk and compliance · executives and CIOsPrerequisites: Know what MTTD and MTTR measure. The page recaps them in one sentence.

Reading mode

An illustrative scenario: the IT department asks for €150,000 a year for an observability project. The CFO does not dispute the technical merits. She asks a single question: “What do we get back and how soon?” This simulator helps answer it honestly, assumptions included.

Think of firefighters. The damage from a fire depends on two durations: the time it takes to notice there is a fire, then the time it takes to put it out. An IT incident is measured the same way. The mean time to detect is called MTTD. The mean time to restore service is called MTTR. An observability project rarely justifies itself. It is justified by what it changes in those two durations, hence in the cost of incidents.

The simulator links four quantities leadership knows or can estimate (the number of incidents, the cost of an hour of downtime, MTTD and MTTR) to a return on investment. Every assumption stays visible and editable.

Incidents
Cost of one hour of downtime (major incident)
Detection and restoration times
Observability investment

Result

Annual incident cost, before
Annual incident cost, after
Annual gain
Annual return on investment
Payback period
Sensitivity: if the improvement is only 50%
Annual incident cost, broken down
  • Detection
  • Restoration
Cumulative over 36 months
  • Cumulative investment cost
  • Cumulative gains

Deliberately simple model: the cost of an incident is proportional to its total duration (MTTD + MTTR, MTTR being counted here from detection); a minor incident costs a fixed fraction of a major one; the investment does not reduce the number of incidents (a cautious assumption). Gains rise linearly during ramp-up. Costs are not discounted. Replace every assumption with your own figures.

  1. Start from the starting case: 24 incidents a year, one in four of them major, MTTD brought down from 30 to 10 minutes and MTTR from 120 to 60 minutes. The annual gain exceeds €290,000 for a cost of €150,000 a year. The return is about 94% while the investment pays back in a little under 10 months.
  2. Look at the “if the improvement is only 50%” tile. The return turns slightly negative: the project only holds if you get more than half of the promised improvement. That is the question to ask before signing.
  3. Set lost revenue to 0, as for an internal service. With only productivity and penalties, the gain no longer covers the cost. The business case depends entirely on what an hour of downtime really costs.
  4. Set the target MTTR to 120 minutes so that only detection time improves. The gain drops to about €73,000. In this case detection accounts for only a fifth of the cost: restoration (diagnosis, correlating signals, written troubleshooting procedures) carries most of the gain. For a service prone to silent failures, detected late, it would be the other way round.

An honest business case does not present a single return figure but a range: the target case, the case where you only get half the improvement, plus the threshold below which the investment does not pay back. The assumptions (hourly cost, share of major incidents, expected gains) must be written down and owned by someone accountable for them. I recommend setting a date one year out as soon as the project is signed, to compare them with the MTTD and MTTR actually measured.

Going further: the business dimension, the bridges between dimensions, the generative AI economic exposure calculator and the error budget simulator.